Literature.md

Automatic Music Transcription: An Overview

Great discussion of SOTA(2019) techniques and challeges for this task. Very useful! http://www.eecs.qmul.ac.uk/~ewerts/publications/2019_BenetosDixonDuanEwert_AutomaticMusicTranscription_IEEE-SPM.pdf

Discussion of CNN feature map sizing for music applications

http://www.jordipons.me/cnn-filter-shapes-discussion/

AMT Example CNN and libraries

Example of auto music transcription for polyphonic piano that has useful python libraries for audio and midi manipulation. It is a fundamentally different ML architecture than our approach (cnn vs rcc) but has many links to relevant signal processing resources. Need to investigate the structure of the example data in this repository. https://github.com/jsleep/wav2mid

Keras SEQ2SEQ implementation

Blogpost on a keras seq-to-seq neural network for machine translation (english -> french). Uses encoder decoder approach which is what we want but data type and shape is very different. https://blog.keras.io/a-ten-minute-introduction-to-sequence-to-sequence-learning-in-keras.html

Study of ML approaches to AMT

Stanford study published on automatic music transcription. Again, concerned with pitch more so than instrument classification. Tries both CNN and RNN architectures but prefers CNN. Useful discussion on data preparation and more sources. http://cs229.stanford.edu/proj2017/final-reports/5242716.pdf

How to interpret midi data for the human

Resource for understanding the structure of midi data in CSV. https://www.fourmilab.ch/webtools/midicsv/